CN114462481A - Personnel safety monitoring method and equipment based on machine learning - Google Patents
Personnel safety monitoring method and equipment based on machine learning Download PDFInfo
- Publication number
- CN114462481A CN114462481A CN202111602916.0A CN202111602916A CN114462481A CN 114462481 A CN114462481 A CN 114462481A CN 202111602916 A CN202111602916 A CN 202111602916A CN 114462481 A CN114462481 A CN 114462481A
- Authority
- CN
- China
- Prior art keywords
- data
- motion
- safety monitoring
- personnel
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Medical Informatics (AREA)
- Alarm Systems (AREA)
- Emergency Alarm Devices (AREA)
Abstract
The invention discloses a personnel safety monitoring method and equipment based on machine learning, belongs to the technical field of safety monitoring, and is used for solving the technical problems that the existing personnel safety monitoring method is not comprehensive in monitoring range and low in safety condition evaluation accuracy of workers. The method comprises the following steps: the human body motion data of the monitored personnel are collected in real time through a plurality of motion sensors; carrying out motion characteristic recognition on the human motion data through the trained motion characteristic recognition model to obtain the motion characteristic data of the monitored personnel; collecting vital sign data of a monitored person, environment data and equipment operation data in real time to obtain a safety monitoring data set; preprocessing the safety monitoring data set and performing clustering analysis to obtain the risk index and the risk index weight of each type of data in the complete safety monitoring data set; and obtaining a danger evaluation value of the monitored personnel according to the danger index and the danger index weight, and further determining whether to give an alarm or not.
Description
Technical Field
The application relates to the technical field of safety monitoring, in particular to a personnel safety monitoring method and equipment based on machine learning.
Background
For workers in dangerous industries such as underground operation, overhead operation, electrical equipment maintenance, chemical engineering and the like, various potential safety hazards are often faced in the working process, for example, situations such as coma, sudden diseases, irregular equipment operation and harm to the workers due to abnormal equipment caused by toxic gas suction can be caused. With the progress of society, various industries are sequentially subjected to informationized construction, so that the monitoring and early warning of human body data are realized, and the safety of workers and working environment is guaranteed.
However, most of the existing methods collect the physical sign data of the staff through devices such as sensors and the like, and monitor and early warn. The data acquired by the method are not comprehensive enough, only the physical conditions of the workers can be monitored, and the potential safety hazards faced by the workers cannot be comprehensively analyzed. The collecting device often collects dead corners, and the monitoring device also depends on the degree of seriousness of personnel, so that the safety monitoring is not in place.
Disclosure of Invention
The embodiment of the application provides a personnel safety monitoring method and equipment based on machine learning, which are used for solving the following technical problems: the existing personnel safety monitoring method is not comprehensive in monitoring range and low in safety condition evaluation accuracy for workers.
The embodiment of the application adopts the following technical scheme:
in one aspect, an embodiment of the present application provides a method for monitoring personnel safety based on machine learning, where the method includes: the human body motion data of the monitored personnel are collected in real time through a plurality of motion sensors; the motion sensor is arranged at a key motion position on the wearable device and at least comprises an acceleration sensor and a gyroscope sensor; based on a preset time interval, carrying out motion characteristic recognition on the human motion data through a trained motion characteristic recognition model to obtain the motion characteristic data of the monitored personnel; collecting vital sign data of the monitored person and the environment data in real time through various sensors mounted on the wearable equipment; acquiring equipment operation data in real time through data acquisition equipment installed in the industrial equipment; combining the motion characteristic data, the vital sign data, the environment data and the equipment operation data into a safety monitoring data set; preprocessing various data in the safety monitoring data set respectively to obtain a complete safety monitoring data set; performing cluster analysis on the complete safety monitoring data set to obtain the risk index and the risk index weight of each type of data in the complete safety monitoring data set; and obtaining a danger evaluation value of the monitored personnel according to the danger index and the danger index weight, and further determining whether to give an alarm or not.
The embodiment of the application acquires the human motion data and the vital sign data of the monitored personnel through the wearable device, and the environment data and the operation data of the equipment operated by the monitored personnel can comprehensively analyze the potential safety hazard faced by the monitored personnel from the aspects of the physical health, the environment, the action, the equipment operation condition beside, and the like of the monitored personnel.
In a possible implementation manner, before the motion feature recognition is performed on the human motion data through the trained motion feature recognition model, the method further includes: connecting the first preset number of convolution layers and one pooling layer in series to obtain a convolution block; stacking a second preset number of convolution blocks to obtain a convolution network; wherein, the value ranges of the first preset quantity and the second preset quantity are both [1,10 ]; combining the convolution network with a double-layer long and short term memory network (LSTM), a full connection layer and a plurality of logistic regression layers to obtain the motion characteristic identification model; training and testing the motion characteristic recognition model through a human motion recognition HAR data set to obtain a test accuracy value; if the test precision value is smaller than a first preset threshold value, the values of the first preset number and the second preset number are adjusted randomly until the test precision reaches the first preset threshold value.
The embodiment of the application connects a plurality of convolution layers with smaller size in series to replace a larger convolution layer, on one hand, richer nonlinear transformation is provided for the model, and on the other hand, the depth of the model can be increased to extract high-dimensional features. And the pooling layer realizes feature dimension reduction, controls the over-fitting risk and keeps the translation invariance. A plurality of convolution layers connected in series and a pooling layer are defined as a convolution block, and the model can be freely stacked with a plurality of convolution blocks, so that the depth of the model can be effectively increased. And the structure information of the data is further mined by combining a long-term and short-term memory network (LSTM), so that the motion recognition of the human body can be realized more accurately and comprehensively.
In a feasible implementation manner, based on a preset time interval, the motion feature recognition is performed on the human motion data through a trained motion feature recognition model to obtain the motion feature data of the monitored person, which specifically includes: sequencing the human motion data acquired by each motion sensor in the current time interval according to the acquisition time to obtain a corresponding human motion data sequence; aligning and arranging each human body motion data sequence to obtain a motion data matrix of the monitored personnel; each row of elements in the motion data matrix corresponds to a human motion data sequence acquired by a motion sensor; inputting the motion data matrix into the motion characteristic identification model, and performing characteristic identification on the motion data matrix through the convolution network to obtain a plurality of one-dimensional characteristic segments; splicing the plurality of one-dimensional feature segments through the double-layer long-short term memory network LSTM to obtain a one-dimensional feature vector; and processing the one-dimensional feature vectors through the full-connection layer, and inputting output results into the multiple logistic regression layer for multi-classification regression processing to obtain the motion feature data.
In a feasible implementation manner, preprocessing each type of data in the security monitoring data set to obtain a complete security monitoring data set includes: determining a maximum threshold value and a minimum threshold value of each type of data according to the types of the data; for each type of data in the safety monitoring data set, determining data with a data value equal to 0 in the current time interval as blank data; determining data with a data value greater than the maximum threshold value or less than the minimum threshold value in the current time interval as error data; filling the blank data according to the data value of the position corresponding to the blank data in the complete monitoring data set in the previous time interval so as to complete various data in the current time interval; and replacing the error data according to the data value of the position corresponding to the error data in the complete monitoring data set in the previous time interval so as to correct various data in the current time interval and obtain the complete safety monitoring data set in the current time interval.
According to the embodiment of the application, the safety monitoring data are more complete and accurate through completion and correction of the safety monitoring data, and the calculation accuracy of the danger evaluation value is improved.
In a feasible implementation manner, performing cluster analysis on the preprocessed security monitoring data set to obtain a risk index and a risk index weight of each data in the security monitoring data set, specifically including: respectively converting the preprocessed vital sign data, the preprocessed environmental data, the preprocessed equipment operation data and the preprocessed motion characteristic data into corresponding data matrixes according to data types; according to
Obtaining cluster analysis fitness M corresponding to the mth data; wherein, PmData matrix for data of m-th class, max (P)m) Is a matrix PmMinimum of middle element, min (P)m) Is a matrix PmThe maximum value of the element(s) in (1), lambda is a clustering factor, and q is an offset difference; according to
Obtaining a danger index rho corresponding to the mth class of data in the safety monitoring data setm(ii) a According to
Obtaining a danger index weight V corresponding to the mth class of data in the safety monitoring data setm(ii) a Wherein the content of the first and second substances,h is the importance degree of the mth type data, and the importance degree is specified by people.
In a possible implementation manner, obtaining the risk evaluation value of the monitored person according to the risk index and the risk index weight specifically includes: according to
Obtaining a danger evaluation value R of the monitored personnel; wherein L is the total number of the data matrix.
In a possible embodiment, after obtaining the risk assessment value of the monitored person according to the risk index and the risk index weight, the method further includes: acquiring the position information of all monitored personnel, and sending the position information, the safety monitoring data set and the danger evaluation value of each monitored personnel to the visual monitoring equipment; generating corresponding icons at corresponding positions in a working scene two-dimensional map in the visual monitoring equipment according to the working types and the position information of monitored personnel; and reading data in the safety monitoring data set, and linking the data with the icon in a tag form, so that the content of the tag is displayed when a cursor passes through the icon.
According to the embodiment of the application, the positions and the safety states of the monitored personnel in the working scene are clear at a glance by performing personalized visual processing on the positions and the safety monitoring data of all the monitored personnel, and the monitoring personnel can conveniently monitor the health conditions of the monitored personnel.
In a possible embodiment, after reading the data in the security monitoring data set and linking with the icon in the form of a tag, so that the content of the tag is displayed when a cursor passes over the icon, the method further comprises: determining that the monitored personnel are threatened by safety under the condition that the danger evaluation value exceeds a second preset threshold value; determining the safety early warning level of the monitored personnel based on the value of the danger evaluation value exceeding the second preset threshold value; the safety early warning level specifically comprises a low-level safety early warning, a middle-level safety early warning and an emergency safety early warning; based on the safety early warning level, converting the corresponding icon in the visual monitoring equipment into a corresponding display state; responding to the operation that the visual monitoring equipment watchman clicks the early warning processing staff, or distributing the processing tasks to the early warning processing staff after the automatic processing time is reached.
In one possible embodiment, after responding to an operation of a visual monitoring device attendant clicking on an early warning processing person or allocating a processing task to the early warning processing person after reaching an automatic processing time, the method further includes: after the monitored personnel are determined to be threatened safely, determining the assignment condition of early warning processing personnel once every preset time; wherein the preset time is less than the automatic processing time; and if no assignment operation is detected within the preset time, the safety early warning level is adjusted up to one level until the emergency early warning is adjusted.
The embodiment of the application provides a perfect early warning and early warning processing mechanism, and can ensure that the monitored personnel can be timely processed or rescued after danger occurs.
On the other hand, this application embodiment still provides a personnel safety monitoring equipment based on machine learning, and equipment includes: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a method for machine learning based personnel security monitoring according to any of the above embodiments.
The personnel safety monitoring equipment based on machine learning provided by the embodiment of the application can influence the safety of the staff by acquiring factor data of various aspects, and obtains the danger evaluation value of the monitored staff through a comprehensive and accurate model and a calculation method, thereby realizing comprehensive monitoring of the safety of the monitored staff and ensuring the life safety of high-risk staff.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings can be obtained by those skilled in the art without creative efforts. In the drawings:
fig. 1 is a flowchart of a personnel safety monitoring method based on machine learning according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of a machine learning-based personal safety monitoring device according to an embodiment of the present application.
Detailed Description
In order to make those skilled in the art better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any inventive step based on the embodiments of the present disclosure, shall fall within the scope of protection of the present application.
The embodiment of the application provides a personnel safety monitoring method based on machine learning, and as shown in fig. 1, the method specifically comprises the following steps of S101-S107:
s101, collecting human body motion data of a monitored person in real time through a plurality of motion sensors.
Specifically, the personnel safety monitoring method based on machine learning is realized on the basis of designed wearable equipment provided with various sensors. The application firstly provides wearable equipment, which comprises a waistcoat, a helmet, a wrist pad and a knee pad; wherein, four groups of motion sensors are respectively arranged on the wrist pad and the knee pad, and each group of motion sensor comprises an acceleration sensor and a gyroscope sensor. The acceleration sensor is used for collecting the motion acceleration of the hand or the leg, and the gyroscope sensor is used for collecting the position of the hand or the leg.
Further, the human body motion data of the monitored person wearing the equipment is collected in real time through the four groups of motion sensors; wherein the human motion data comprises the acceleration and the position of two hands and two legs of the monitored person.
And S102, constructing and training a motion characteristic recognition model.
Specifically, a first preset number of convolution layers and a pooling layer are connected in series to obtain a convolution block, and then a second preset number of convolution blocks are stacked to obtain a convolution network. Wherein, the value ranges of the first preset number and the second preset number are both [1,10 ]. Combining the convolution network with a double-layer Long-Short Term Memory (LSTM), a full connection layer and a multi-item logistic regression layer to obtain a motion characteristic identification model; training and testing the motion characteristic recognition model through a human motion recognition HAR data set to obtain a test accuracy value; and if the test precision value is smaller than the first preset threshold value, randomly adjusting the values of the first preset quantity and the second preset quantity until the test precision reaches the first preset threshold value.
The embodiment of the application connects a plurality of convolution layers with smaller size in series to replace a larger convolution layer, on one hand, richer nonlinear transformation is provided for the model, and on the other hand, the depth of the model can be increased to extract high-dimensional features. And the pooling layer realizes feature dimension reduction, controls the over-fitting risk and keeps the translation invariance. A plurality of convolution layers connected in series and a pooling layer are defined as a convolution block, and the model can be freely stacked with a plurality of convolution blocks, so that the depth of the model can be effectively increased. And the structure information of the data is further mined by combining a long-term and short-term memory network (LSTM), so that the motion recognition of the human body can be realized more accurately and comprehensively.
S103, based on a preset time interval, carrying out motion characteristic recognition on the human motion data through the trained motion characteristic recognition model to obtain the motion characteristic data of the monitored personnel.
Specifically, the human motion data collected by each motion sensor in the current time interval are sorted according to the collection time to obtain a corresponding human motion data sequence. Then aligning and arranging each human body motion data sequence to obtain a motion data matrix of the monitored personnel. Each row of elements in the motion data matrix corresponds to a human motion data sequence acquired by a motion sensor.
In one embodiment, if the hand acceleration data acquired by the first acceleration sensor in the current time interval is sorted according to the acquisition time, the data sequence is a human motion data sequence: t is1＝[a1，a2，a3，a4](ii) a The human motion data sequence of the second acceleration sensor is as follows: t is2＝[b1，b2，b3，b4](ii) a The human motion data sequence of the third acceleration sensor is as follows: t is3＝[c1，c2，c3，c4](ii) a The human motion data sequence of the fourth acceleration sensor is as follows: t is4＝[d1，d2，d3，d4](ii) a The human motion data sequence of the first gyroscope sensor is as follows: t is5＝[e1，e2，e3，e4](ii) a The human motion data sequence of the second gyroscope sensor is as follows: t is6＝[f1，f2，f3，f4](ii) a The human motion data sequence of the third gyro sensor is:
T7＝[g1，g2，g3，g4](ii) a The human motion data sequence of the fourth gyroscope sensor is as follows:
T8＝[h1，h2，h3，h4](ii) a The motion data matrix of the monitored person in the current time interval is then:
and further, inputting the motion data matrix into a motion characteristic identification model, and performing characteristic identification on the motion data matrix through a convolution network to obtain a plurality of one-dimensional characteristic segments. And then splicing a plurality of one-dimensional characteristic segments through a double-layer long-short term memory network (LSTM) to obtain a one-dimensional characteristic vector. And then, processing the one-dimensional feature vectors through the full-connection layer, and inputting the output result into a multi-directional logistic regression layer for multi-classification regression processing to obtain motion feature data.
And S104, acquiring a safety monitoring data set.
Specifically, vital sign data of a monitored person and environment data of the monitored person are collected in real time through various sensors mounted on wearable equipment; and the data acquisition equipment installed in the industrial equipment is used for acquiring the operation data of the equipment in real time.
In one embodiment, the vital sign data includes heart rate, pressure, blood glucose, blood pressure, blood oxygen, body temperature, etc. data, which are respectively collected by existing sensors. The environmental data comprise ventilation volume, oxygen concentration and special gas (such as gas) concentration, and are respectively collected through the existing sensors. The equipment operation data comprises voltage, current, motion temperature and other data during equipment operation, and are acquired through the existing sensors respectively.
Further, the motion characteristic data, the vital sign data, the environment data and the equipment operation data are combined into a safety monitoring data set.
And S105, preprocessing various data in the safety monitoring data set respectively to obtain a complete safety monitoring data set.
Specifically, the safety monitoring data set often contains some blank data or error data, and the direct use of the safety monitoring data set for risk evaluation may result in an error or a large error of the calculation result, so that the safety monitoring data set is preprocessed before calculation, and the specific method includes:
and determining the maximum threshold and the minimum threshold of each type of data according to the types of the data. For various types of data in the safety monitoring data set, determining the data with the data value equal to 0 in the current time interval as blank data; and determining the data with the data value larger than the maximum threshold value or smaller than the minimum threshold value in the current time interval as error data. And then filling the blank data according to the data values of the positions corresponding to the blank data in the complete monitoring data set in the previous time interval so as to fill various data in the current time interval. And replacing the error data according to the data value of the position corresponding to the error data in the complete monitoring data set in the previous time interval so as to correct various data in the current time interval and obtain the complete safety monitoring data set in the current time interval.
Taking heart rate data as an example, it can be determined according to medical general knowledge that the heart rate of a person can reach 200 times/second at the fastest time and can reach 40 times/second at the lowest time, and therefore if the monitored heart rate is not within the range of 40-200 times/second, data errors are likely to occur. Therefore, data with the heart rate value not within the range of 40-200 times/second in the current time interval is determined as error data, the sequence of the error data in the heart rate data in the current time interval is determined, the heart rate value of the corresponding sequence in the previous time interval is inquired, and the error data is replaced. For example, if the 3 rd heart rate value is 300/s in the current time interval, the 3 rd heart rate value is error data, and the 3 rd heart rate value in the previous time interval is queried and replaced with the 3 rd heart rate value in the current time interval. It should be noted that since the complete monitoring data set in each time interval is preprocessed, the heart rate values in the corresponding sequence in the previous time interval are not necessarily error data or blank data. Other types of data may be provided with maximum and minimum thresholds based on common sense or experience, and are not listed in this application.
And S106, carrying out cluster analysis on the complete safety monitoring data set to obtain the risk index and the risk index weight of each type of data in the complete safety monitoring data set.
In particular, according to
Obtaining cluster analysis fitness M corresponding to the mth data; wherein, PmData matrix for data of m-th class, max (P)m) Is a matrix PmMinimum value of element (1)， min(Pm) Is a matrix PmThe maximum value of the element(s) in (b), λ is the clustering factor, and q is the offset difference.
Further in accordance with
Obtaining a danger index rho corresponding to the mth class of data in the safety monitoring data setm。
Further in accordance with
Obtaining the danger index weight V corresponding to the mth data in the safety monitoring data setm(ii) a Wherein, H is the importance degree of the mth data, and the importance degree can be manually specified according to the grading of experts.
And S107, obtaining a danger evaluation value of the monitored personnel according to the danger index and the danger index weight, and further determining whether to give an alarm or not.
In particular, according to
Obtaining a danger evaluation value R of the monitored personnel; wherein L is the total number of data matrices. And then acquiring the position information of all monitored personnel, and sending the position information, the safety monitoring data set and the danger evaluation value of each monitored personnel to the visual monitoring equipment.
Furthermore, in a two-dimensional map of a working scene in the visual monitoring equipment, corresponding icons are generated at corresponding positions according to the working types and the position information of the monitored personnel. And then reading the data in the safety monitoring data set, and linking the data with the icon in the form of a label so as to display the content of the label when the cursor passes through the icon.
Further, in the case that the danger evaluation value exceeds a second preset threshold value, it is determined that the monitored person is threatened by safety. And determining the safety early warning level of the monitored personnel based on the value of the danger evaluation value exceeding a second preset threshold value. The safety early warning level specifically comprises low-level safety early warning, medium-level safety early warning and emergency safety early warning.
In one embodiment, if the risk evaluation value is a and the second preset threshold is B, the safety precaution level is a low-level safety precaution if a exceeds B by a value greater than or equal to the low-level early warning threshold and less than the medium-level early warning threshold. And if the value of A exceeding B is greater than or equal to the middle-level early warning threshold and smaller than the high-level early warning threshold, the safety early warning level is middle-level safety early warning. And if the numerical value of A exceeding B is more than or equal to the advanced early warning threshold value, the safety early warning level is advanced safety early warning.
Further, based on the safety early warning level, the corresponding icon in the visual monitoring device is converted into the corresponding display state. Responding to the operation that the visual monitoring equipment watchman clicks the early warning processing staff, or distributing the processing tasks to the early warning processing staff after the automatic processing time is reached. After the monitored personnel are determined to be threatened safely, the assignment condition of the early warning processing personnel is determined once every preset time. Wherein the preset time is less than the automatic processing time. And if no assignment operation is detected within the preset time, the safety early warning level is adjusted up to one level until the emergency early warning is adjusted.
In addition, an embodiment of the present application further provides a device for monitoring personal safety based on machine learning, as shown in fig. 2, the device for monitoring personal safety 200 based on machine learning specifically includes:
at least one processor 201; and a memory 202 communicatively coupled to the at least one processor 201; wherein the memory 202 stores instructions executable by the at least one processor 201 to enable the at least one processor 201 to:
the human body motion data of the monitored personnel are collected in real time through a plurality of motion sensors; the motion sensor is arranged at a key motion position on the wearable equipment and at least comprises an acceleration sensor and a gyroscope sensor;
based on a preset time interval, carrying out motion characteristic recognition on human motion data through a trained motion characteristic recognition model to obtain motion characteristic data of a monitored person;
the method comprises the steps that vital sign data of a monitored person and environment data of the monitored person are collected in real time through various sensors mounted on wearable equipment; acquiring equipment operation data in real time through data acquisition equipment installed in the industrial equipment;
combining the motion characteristic data, the vital sign data, the environment data and the equipment operation data into a safety monitoring data set;
preprocessing various data in the safety monitoring data set respectively to obtain a complete safety monitoring data set;
performing cluster analysis on the complete safety monitoring data set to obtain the risk index and the risk index weight of each type of data in the complete safety monitoring data set;
and obtaining a danger evaluation value of the monitored personnel according to the danger index and the danger index weight, and further determining whether to give an alarm or not.
The embodiments in the present application are described in a progressive manner, and the same and similar parts among the embodiments can be referred to each other, and each embodiment focuses on differences from other embodiments. In particular, for the apparatus embodiment, since it is substantially similar to the method embodiment, the description is relatively simple, and reference may be made to the partial description of the method embodiment for relevant points.
The foregoing description of specific embodiments of the present application has been presented. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The above description is only an example of the present application and is not intended to limit the present application. Various modifications and changes may occur to those skilled in the art to which the embodiments of the present application pertain. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of the present application should be included in the scope of the claims of the present application.
Claims (10)
1. A personnel safety monitoring method based on machine learning is characterized by comprising the following steps:
the human body motion data of the monitored personnel are collected in real time through a plurality of motion sensors; the motion sensor is arranged at a key motion position on the wearable device and at least comprises an acceleration sensor and a gyroscope sensor;
based on a preset time interval, carrying out motion characteristic recognition on the human motion data through a trained motion characteristic recognition model to obtain the motion characteristic data of the monitored personnel;
collecting vital sign data of the monitored person and the environment data in real time through various sensors mounted on the wearable equipment; acquiring equipment operation data in real time through data acquisition equipment installed in the industrial equipment;
combining the motion characteristic data, the vital sign data, the environment data and the equipment operation data into a safety monitoring data set;
preprocessing various data in the safety monitoring data set respectively to obtain a complete safety monitoring data set;
performing cluster analysis on the complete safety monitoring data set to obtain the risk index and the risk index weight of each type of data in the complete safety monitoring data set;
and obtaining a danger evaluation value of the monitored personnel according to the danger index and the danger index weight, and further determining whether to give an alarm or not.
2. The personnel safety monitoring method based on machine learning of claim 1, wherein before the human motion data is subjected to motion feature recognition through a trained motion feature recognition model, the method further comprises:
connecting the first preset number of convolution layers and one pooling layer in series to obtain a convolution block;
stacking a second preset number of convolution blocks to obtain a convolution network; wherein, the value ranges of the first preset quantity and the second preset quantity are both [1,10 ];
combining the convolution network with a double-layer long and short term memory network (LSTM), a full connection layer and a plurality of logistic regression layers to obtain the motion characteristic identification model;
training and testing the motion characteristic recognition model through a human motion recognition HAR data set to obtain a test accuracy value;
if the test precision value is smaller than a first preset threshold value, the values of the first preset number and the second preset number are adjusted randomly until the test precision reaches the first preset threshold value.
3. The personnel safety monitoring method based on machine learning according to claim 2, characterized in that, based on a preset time interval, the human body movement data is subjected to movement feature recognition through a trained movement feature recognition model to obtain the movement feature data of the monitored personnel, specifically comprising:
sequencing the human motion data acquired by each motion sensor in the current time interval according to the acquisition time to obtain a corresponding human motion data sequence;
aligning and arranging each human body motion data sequence to obtain a motion data matrix of the monitored personnel; each row of elements in the motion data matrix corresponds to a human motion data sequence acquired by a motion sensor;
inputting the motion data matrix into the motion characteristic identification model, and performing characteristic identification on the motion data matrix through the convolution network to obtain a plurality of one-dimensional characteristic segments;
splicing the plurality of one-dimensional feature segments through the double-layer long-short term memory network LSTM to obtain a one-dimensional feature vector;
and processing the one-dimensional feature vectors through the full-connection layer, and inputting output results into the multiple logistic regression layer for multi-classification regression processing to obtain the motion feature data.
4. The personnel safety monitoring method based on machine learning of claim 1, characterized in that, preprocessing is performed on each kind of data in the safety monitoring data set respectively to obtain a complete safety monitoring data set, specifically comprising:
determining a maximum threshold value and a minimum threshold value of each type of data according to the types of the data;
for each type of data in the safety monitoring data set, determining data with a data value equal to 0 in the current time interval as blank data; determining data with a data value greater than the maximum threshold value or less than the minimum threshold value in the current time interval as error data;
filling the blank data according to the data value of the position corresponding to the blank data in the complete monitoring data set in the previous time interval so as to complete various data in the current time interval;
and replacing the error data according to the data value of the position corresponding to the error data in the complete monitoring data set in the previous time interval so as to correct various data in the current time interval and obtain the complete safety monitoring data set in the current time interval.
5. The personnel safety monitoring method based on machine learning of claim 1, wherein the cluster analysis is performed on the preprocessed safety monitoring data set to obtain a risk index and a risk index weight of each data in the safety monitoring data set, and specifically comprises:
respectively converting the preprocessed vital sign data, the preprocessed environmental data, the preprocessed equipment operation data and the preprocessed motion characteristic data into corresponding data matrixes according to data types;
according to
Obtaining cluster analysis fitness M corresponding to the mth data; wherein, PmData matrix for data of m-th class, max (P)m) Is a matrix PmMinimum of middle element, min (P)m) Is a matrix PmThe maximum value of the element(s) in (1), lambda is a clustering factor, and q is an offset difference;
according to
Obtaining a danger index rho corresponding to the mth class of data in the safety monitoring data setm；
According to
Obtaining a danger index weight V corresponding to the mth class of data in the safety monitoring data setm(ii) a And H is the importance degree of the mth type data, and the importance degree is specified by people.
6. The machine learning-based personnel safety monitoring method according to claim 5, wherein obtaining the risk evaluation value of the monitored personnel according to the risk index and the risk index weight specifically comprises:
according to
Obtaining a danger evaluation value R of the monitored personnel; wherein L is the total number of the data matrix.
7. The machine learning-based personnel safety monitoring method according to claim 1, wherein after obtaining the risk evaluation value of the monitored personnel according to the risk index and the risk index weight, the method further comprises:
acquiring the position information of all monitored personnel, and sending the position information, the safety monitoring data set and the danger evaluation value of each monitored personnel to the visual monitoring equipment;
generating corresponding icons at corresponding positions in a working scene two-dimensional map in the visual monitoring equipment according to the working types and the position information of monitored personnel;
and reading data in the safety monitoring data set, and linking the data with the icon in a tag form, so that the content of the tag is displayed when a cursor passes through the icon.
8. The machine learning-based personnel security monitoring method of claim 7, wherein after reading the data in the security monitoring data set and linking with the icon in the form of a tag to display the content of the tag when a cursor passes over the icon, the method further comprises:
determining that the monitored personnel are threatened by safety under the condition that the danger evaluation value exceeds a second preset threshold value;
determining the safety early warning level of the monitored personnel based on the value of the danger evaluation value exceeding the second preset threshold value; the safety early warning level specifically comprises a low-level safety early warning, a middle-level safety early warning and an emergency safety early warning;
based on the safety early warning level, converting the corresponding icon in the visual monitoring equipment into a corresponding display state;
responding to the operation that the visual monitoring equipment watchman clicks the early warning processing staff, or distributing the processing tasks to the early warning processing staff after the automatic processing time is reached.
9. The machine learning-based personnel safety monitoring method according to claim 8, wherein after responding to the operation of a visual monitoring device attendant clicking on an early warning processing personnel or after reaching an automatic processing time and allocating a processing task to the early warning processing personnel, the method further comprises:
after the monitored personnel are determined to be threatened safely, determining the assignment condition of early warning processing personnel once every preset time; wherein the preset time is less than the automatic processing time;
and if no assignment operation is detected within the preset time, the safety early warning level is adjusted up to one level until the emergency early warning is adjusted.
10. A machine learning based personal safety monitoring device, the device comprising:
at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein, the first and the second end of the pipe are connected with each other,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform a method for machine learning based personnel security monitoring according to any of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111602916.0A CN114462481A (en) | 2021-12-24 | 2021-12-24 | Personnel safety monitoring method and equipment based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111602916.0A CN114462481A (en) | 2021-12-24 | 2021-12-24 | Personnel safety monitoring method and equipment based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114462481A true CN114462481A (en) | 2022-05-10 |
Family
ID=81408597
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111602916.0A Pending CN114462481A (en) | 2021-12-24 | 2021-12-24 | Personnel safety monitoring method and equipment based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114462481A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115243404A (en) * | 2022-07-07 | 2022-10-25 | 北京比福特科技发展有限公司 | Electronic T-card terminal control system and method |
| CN117560465A (en) * | 2023-03-07 | 2024-02-13 | 河北地质大学 | Mine safety monitoring system and method based on Internet |
-
2021
- 2021-12-24 CN CN202111602916.0A patent/CN114462481A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115243404A (en) * | 2022-07-07 | 2022-10-25 | 北京比福特科技发展有限公司 | Electronic T-card terminal control system and method |
| CN115243404B (en) * | 2022-07-07 | 2024-01-16 | 北京比福特科技发展有限公司 | Electronic T card terminal control system and method |
| CN117560465A (en) * | 2023-03-07 | 2024-02-13 | 河北地质大学 | Mine safety monitoring system and method based on Internet |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Antwi-Afari et al. | Wearable insole pressure system for automated detection and classification of awkward working postures in construction workers | |
| CN114462481A (en) | Personnel safety monitoring method and equipment based on machine learning | |
| Antwi-Afari et al. | Construction activity recognition and ergonomic risk assessment using a wearable insole pressure system | |
| Kucukoglu et al. | Application of the artificial neural network method to detect defective assembling processes by using a wearable technology | |
| Antwi-Afari et al. | Deep learning-based networks for automated recognition and classification of awkward working postures in construction using wearable insole sensor data | |
| Yang et al. | Automated detection of near-miss fall incidents in iron workers using inertial measurement units | |
| CN106580282A (en) | Human body health monitoring device, system and method | |
| CN111583592B (en) | Experimental environment safety early warning method based on multidimensional convolution neural network | |
| CN117438024B (en) | Intelligent acquisition and analysis system and method for acute diagnosis patient sign data | |
| CN112766429B (en) | Method, device, computer equipment and medium for anomaly detection | |
| CN109541022A (en) | A kind of bridge structure crack health monitoring analysis method | |
| WO2022026658A1 (en) | System and method for monitoring musculoskeletal loading and applications of same | |
| CN104269025A (en) | Wearable type single-node feature and position selecting method for monitoring outdoor tumble | |
| CN111403031A (en) | Physique test data acquisition and detection system and method | |
| CN113050478A (en) | Method for monitoring and early warning operation state of transformer substation constructor | |
| CN112911017A (en) | Operating personnel safety state risk management and control system based on intelligent Internet | |
| AU2024201905A1 (en) | Method and System for Monitoring Core Body Movements | |
| CN114781547A (en) | Method and system for state monitoring and fault diagnosis of chemical equipment | |
| Hong et al. | Monitoring safety behaviors of scaffolding workers using Gramian angular field convolution neural network based on IMU sensing data | |
| US11704615B2 (en) | Risk assessment apparatus and related methods | |
| CN113034499B (en) | Slope surface danger area prediction method, device, equipment and readable storage medium | |
| CN116019429A (en) | Health monitoring method, device, equipment and storage medium based on physiological index | |
| Eltanani et al. | K-Nearest Neighbor Algorithm: Proposed Solution for Human Gait Data Classification | |
| CN114371076A (en) | Method and system for testing stress value of workpiece, electronic equipment and storage medium | |
| CN116434478B (en) | Intelligent early warning response method, device and system for geological disasters |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |